Campaign 0: shakedown.

[brettviren]

Goal: run something with SPNG to shake out problems in code, configuration and infrastructure the development has yet to catch. Sim only and output need not necessarily be useful other than it should demonstrate no glaring bugs. Include CPU-only + CPU+GPU modes.

• Finalize the configuration that allows CPU/GPU choice as Jsonnet TLA's. An OSP vs SPNG variant config is also good. Commit all configuration to repo.
• Define command lines (wire-cell stand-alone) as shell or BATS scripts. Commit these to repo.
• Test command lines locally and show output is "reasonable".
• Tag SPNG repo
• Collect DNNROI model file (wcgpu:/nfs/data/1/abashyal/spng/model_files/Pytorch-UNet/unet-l23-cosmic500-e50.ts) for the container image.
• Rerun winch to make a podman container, dump to oci and sif, test the command lines work locally, publish the images.
• Collect the input dataset over which the campaign will map the command lines. These will be provided outside of the container. One test 55 MB test file holding ADC waveforms has been used. This is maybe enough for shake-down but we could think to add more.
• Write the job submission configs. They should supprot capture and extraction of all data and log files produced.
• Evaluate output, document resource usage, make a table, write some note...
• What else?....

[brettviren]

From Jake:

An input file:

/nfs/data/1/calcuttj/spng_merging2/toolkit/spng/test/tensor_frames_noise_1anode.npz

A command line:

wire-cell -l stdout -L debug -P ../../cfg/ -C elecGain=7.8 -C nanodes=1 \
    -A input_file=tensor_frames_noise_1anode.npz \
    --tla-code do_spng=true \
    -A ts_model_file=/nfs/data/1/abashyal/spng/model_files/Pytorch-UNet/ts-model-2.3/nestedunet-l23-cosmic500-e50.ts \
    wct-framesource_1anode_dnnroi.jsonnet

[brettviren]

Container images are collecting here: https://www.phy.bnl.gov/~bviren/wire-cell/spng/

[brettviren]

Container related details are here: https://github.com/brettviren/winch/blob/master/doc/spng.org#campaign-0